Write a comment about the necessity of using a separate thread

rhpvorderman · rhpvorderman · commit 31694237851c · 2024-03-19T14:25:46.000+01:00
diff --git a/src/xopen/__init__.py b/src/xopen/__init__.py
@@ -249,6 +249,15 @@ def __init__(  # noqa: C901
         assert self.process.stdin is not None
         if "r" in mode:
             self.in_pipe = self.process.stdin
+            # A python subprocess can read and write from pipes, but not from
+            # Python in-memory objects. In order for a program to read from an
+            # in-memory object, a pipe must be created. This pipe must be fed
+            # data from the in-memory object. This must be done in a separate
+            # thread, because IO operations will block when the pipe is full
+            # when writing, or empty when reading. Since the quantity of output
+            # data generated by a certain amount of input data is unknown, the
+            # only way to prevent a blocking application is to write
+            # data continuously to the process stdin on another thread.
             self.in_thread = threading.Thread(target=self._feed_pipe)
             self.in_thread.start()
             self._file: BinaryIO = self.process.stdout  # type: ignore
